OpenAI Sora—the video generation tool announced earlier this year—has been officially launched by OpenAI. Those with the $200/month ChatGPT Pro subscription can now use it to generate FHD, 20-second-long video clips. The text-based models are already quite advanced, so it’s nice to have a change in direction. As per OpenAI, Sora is a foundation for AI that understands and simulates reality and turns it into pixels.
Here’s everything you need to know about Sora, OpenAI’s video generation tool.
Sora represents a leap forward in OpenAI’s toolkit of AI models. Initially introduced in a research preview earlier this year, the tool has evolved significantly with the release of Sora Turbo, which boasts enhanced speed and efficiency. Users can now create videos up to 1080p resolution and 20 seconds long, with options for various aspect ratios, including widescreen, vertical, and square formats. This versatility allows for numerous creative applications, from social media reels to professional presentations.
The interface has also been revamped, making it easier for users to input their ideas. Sora includes features like a storyboard tool that enables precise control over each frame, allowing creators to specify inputs for text, images, and videos seamlessly. Additionally, users can remix existing assets or generate entirely new content from scratch, fostering innovation and collaboration.
OpenAI Sora Limitations and Safety Concerns
Despite its impressive capabilities, OpenAI has been transparent about Sora’s limitations. The current version sometimes produces unrealistic physics and struggles with complex actions over extended durations. Previously, there were concerns that bad actors could use Sora to quickly generate reams of AI-scripted footage on trending topics to attract views and advertising dollars. This could clog YouTube with artificial content finely optimized to game the platform’s recommendation algorithms. Some also worried Sora would usher in a new era of “content farms” that undercut legitimate creators with endless machine-made videos.
In light of growing concerns about the misuse of AI-generated content—particularly regarding deepfakes and misinformation—OpenAI has implemented several safety measures. All videos generated by Sora will include C2PA metadata, clearly marking them as AI-generated. This transparency is complemented by visible watermarks on each video, helping to combat potential misuse in online disinformation campaigns.
Moreover, OpenAI has restricted uploads of human likenesses at launch to mitigate risks associated with deepfakes. These limitations are expected to evolve as the company enhances its safeguards against abuse.
OpenAI Sora Accessibility and Future Plans
While Sora is currently available only to Plus and Pro subscribers in certain regions, OpenAI has plans to expand access in the future. Notably, the tool is not yet available in the UK, Switzerland, or the European Economic Area due to regulatory concerns. OpenAI is actively working with governments in these regions to navigate content laws and enable broader access.
For users interested in using Sora’s capabilities without incurring high costs, the tool offers a free tier that allows for generating up to 50 videos at 480p resolution each month. For those seeking more extensive use, the Pro plan provides ten times the usage limits along with higher resolutions and longer video durations.
A Call for Responsible Use
OpenAI emphasizes that the launch of Sora is not just about technological advancement but also about fostering the responsible use of AI tools. The company aims to give society time to explore Sora’s possibilities while co-developing norms and safeguards that ensure ethical usage as the technology matures.
As Sora enters the public domain, it opens up exciting possibilities for creators across various fields. From filmmakers looking to visualize scripts quickly to marketers aiming to produce engaging content rapidly, Sora can serve as a powerful ally in storytelling.
But for those of you who require textual models for tasks like writing, coding, etc., you don’t need to spend $200/month and you can use GPT-4o, GPT-4o-mini, and models such as Claude 3.5 Sonnet on Bind AI.